Gazprom - University of Toronto



The Corporate Governance Role of the Media:

Evidence from Russia

Alexander Dyck

University of Toronto

Natalya Volchkova

New Economic School and CEFIR

and

Luigi Zingales

University of Chicago, NBER, and CEPR

Abstract

We study the effect of media coverage on corporate governance by focusing on Russia in the period 1999-2002. This setting offers four ideal conditions for such a study: plenty of corporate governance violations, no alternative mechanisms to address them, the presence of an investment fund (the Hermitage) that actively lobbies the international press to shame companies perpetrating those violations, and firms relatively unsophisticated in interacting with the foreign press. We find that Hermitage’s lobbying is effective in increasing the coverage of corporate governance violations in the Anglo-American press. We also find that coverage in the Anglo-American press increases the probability that a corporate governance violation is reversed: one more article increases the probability of reversal by 5 percentage points. This effect is present even when we instrument coverage with an exogenous determinant, i.e. the Hermitage’s portfolio composition at the beginning of the period. The Hermitage’s strategy seems to work in part by impacting Russian companies’ reputation abroad and in part by forcing regulators into action.

.

* Alexander Dyck thanks the Gamma Foundation, the Division of Research, Harvard Business School, and the Rotman School of Management for financial support. Luigi Zingales thanks the Gamma Foundation, the CRSP center, and the George Stigler Center at the University of Chicago for financial support. We thank Beatriz Armendariz, Stefano della Vigna. Andrei Shleifer, Andrei Simonov, Ekaterina Zhuravskaya, and participants and seminars at Dartmouth, Harvard, Stockholm School of Economics and the NBER for very useful comments. We thank Mehmet Beceren and Victor Xin for their research assistance.

In recent years, hedge funds have emerged as some of the most powerful players in corporate governance worldwide. From the dismissal of Deutsche Boerse’s CEO Seifert to McDonalds spin-off of major assets in an IPO, hedge funds have played a crucial role. The Wall Street Journal labeled them the “new leader” on the “list of bogeymen haunting the corporate boardroom.”[1] Among the many tactics hedge funds managers use, the most prominent one is to focus public attention on an underperforming company and shame the CEO to either resign or change policy (Kahan and Rock, 2006).

It is hard to tell, however, whether this public relations campaign is just a smokescreen for more important maneuvers that take place behind the scene or is a crucial ingredient of their battle. Can hedge funds (or shareholders in general) increase the level of coverage received by certain news/companies? And if so, does this coverage have any effect on corporate governance outcomes? These questions are hard to address using U.S. data. Most hedge funds trade in and out of companies very quickly. So it is hard to disentangle whether they are simply good at recognizing that the situation is ripe for change or whether they are indeed an actor of change. Hedge funds in the United States (and in most of Europe) also have access to an array of options to address bad corporate governance (from shareholder’s suits to calling an extraordinary general meeting). So it is hard to tell whether they succeed because of their public relations campaign or because of the power of their legal rights. In addition, the impact of their media campaign can be reduced by countervailing public relations effort exerted by firms.

To overcome these problems we study shareholders’ ability to influence coverage and the impact of this coverage on corporate governance by looking at Russia. Russia presents a useful laboratory setting for this analysis for several reasons. First, during the late 1990s, corporate governance violations in Russia were very extreme, very common, and very visible, providing a wide field of inquiry. Second, in Russia, the standard mechanisms to readdress these violations were either non-existent or completely ineffective (for example, courts were easily corruptible), allowing us to identify whether media have an independent effect on outcomes.

Third, and most important, in Russia, there exists an investment fund (the Hermitage Fund), with extremely low turnover, that consciously played a media strategy after the 1998 Russian crisis. In the words of its chairman Bill Browder: “Our basic approach is to thoroughly research and understand where the corporate malfeasance is taking place and then go to great pains to simplify the story so the average person can understand what is going on.…. We then share the stories with the press. By doing so, we want to inflict real consequences – business, reputational and financial” (Dyck, 2002). Since the Hermitage fund spends resources only when it has money at stake, we can use the Hermitage’s portfolio composition as an instrument for news coverage.

Fourth, Russian managers were just learning about the impact of the press and were unlikely to factor into their decisions the reputational cost the media could impose.

Last but not least, in Russia there was a major regime shift (at the time of the Russian default), when the level of corporate governance violations exploded. This regime shift makes it unlikely that the pre-Russia default stake of the Hermitage (which we use as an instrument) was chosen with a media strategy in mind, eliminating the risk of reverse causality.

Besides its role as an ideal laboratory setting, the study of alternative mechanisms of corporate governance in an emerging market like Russia is of independent interest. The fraction of pension money invested in emerging markets with unformed legal systems (like China) is rapidly growing. But Western investors often find themselves at a loss in these markets, where most of the U.S.–type of institutional checks and balances do not work. Hence, our study of an effective alternative corporate governance mechanism can be of great practical interest.

To identify a sample of potential corporate governance violations we exploit the fact that a prominent Russian investment bank, Troika Dialog, produced a weekly publication, between 1998 and 2002, that highlighted all the corporate actions that, in their view, have the potential to severely impact outside investors’ rights. This definition of potential violation does not necessarily imply that any Russian law was infringed.[2] How to judge, for instance, Tomskneft’s dilutive equity issue in 1999? The issue was approved by shareholders present at the meeting. But very few were able to be present because the day of the meeting the company announced that the venue had been transferred to a new distant location that shareholders couldn’t possibly reach in time to vote on the proposal.

We refine this list by eliminating repeated events and minor violations (like a delay in financial reporting). We then study how much coverage each of these violations received and whether they were stopped or somehow readdressed.

Not surprisingly, we find that the magnitude of the violation (which we proxy by the potential loss caused by the announced decision) increases the extent it is covered in the Anglo-American media. We also find that, controlling for the severity of the violation, companies receiving more coverage in normal periods (and thus more newsworthy) command more attention. Even controlling for these factors, however, we find that the presence of the Hermitage fund among its shareholders increases the amount of coverage a corporate governance violation receives. This correlation does not appear to be due to the Hermitage fund’s ability to pick newsworthy companies, since the effect is present even when we use the Hermitage Fund’s stake in companies at the beginning of the period (end of 1998).

We then test whether news coverage in the Russian and prominent English language press surrounding and following the revelation of this potential violation is correlated with the eventual outcome. We find that the bad corporate governance decision is reverted following an increase in coverage of the event in Anglo-American newspapers. More importantly, the probability of this reversal is significantly affected by media coverage, even after controlling for other potential determinants of the outcome, such as the degree of foreign ownership and the involvement of international organizations such as the European Bank of Reconstruction and Development (EBRD). By contrast, exposure in the local press has no impact.

One explanation for the irrelevance of domestic newspapers is lack of credibility. Another is that shaming only works if the audience shares the same set of values. If diluting minority shareholders is not perceived as terrible by Russian businessmen, then shaming cannot work. To separate these effects, we use a Russian-language publication called Vedemosti. Since this publication is a joint venture between the Wall Street Journal and the Financial Times, it has credibility similar to that of its owners. But being in Russian, it only reaches Russian businessmen and politicians. Our finding that coverage by Vedemosti has no significant effect suggests that in Russia the only shaming that works is the one in front of the international business community.

So exposure of corporate governance violations in the international press seems to promote some readdress. This evidence hardly proves that the press was an instrument of change, let alone that hedge funds were the force behind this change. An egregious corporate governance violation is more likely to be covered by newspapers regardless of any effort by hedge fund managers. And such an egregious violation is also more likely to generate a reaction. To attempt to disentangle these effects, we instrument foreign press coverage with the Hermitage Fund’s stake in companies at the end of 1998. Since the Hermitage fund will only spend resources in lobbying the press if it has some skin in the game, Hermitage’s stake can be considered a good measure of the exogenous component in news coverage. When we instrument coverage with this exogenous determinant, its estimated impact on outcome does not change, suggesting this link might be causal.

As with any instrument, there may be concern about whether it is truly exogenous. To address these concerns we gather additional evidence, which all point in the direction of causality flowing from news to outcome. In particular, we are able to trace back the mechanism that allows the Hermitage fund to influence the publication of news.

Our estimate of the economic impact of media pressure is large. One standard deviation increase in coverage increases the probability of reversal by 14 percentage points (a 49 percent increase in the average sample probability). One single article more in the Anglo American press buys a 5 percentage point increase in the probability of reversal (equal to a 18 percent increase with respect to the sample average). Since the average corporate governance violation had the potential to dilute the value of equity by 57 percent and the average (median) company had a book value of equity equal to $1,417 million (114 million) by the end of our sample period, then the value of an extra article published in the Wall Street Journal or the Financial Times is $ 40.4 million (3.3 million).[3] If we restrict our attention to those firms with enough trading in their stock to produce a reliable estimate of market value, our estimates are even larger since the average (median) firm had equity value of $2,600 million (288 million), and the corresponding value of an extra article is $ 74.2 million (8.2 million ).[4]

This estimate represents the impact that foreign media can have in Russia. In countries, like the United States, where pro-shareholder values are widely shared among the business community, the impact of media on corporate governance outcomes is likely to be even stronger (and since firms are larger, its value effect even bigger). Not surprisingly, publication of much milder violations (such as the excessive compensation of the former NYSE chairman Richard Grasso) lead to immediate firing or resignations.

Our estimates also suggest that with limited resources the Hermitage was able to double the coverage of an event. This magnitude seems more specific to a developing country like Russia. In more developed countries a fund like Hermitage may find its impact reduced by the countervailing lobbying efforts exercised by the companies targeted. That the equilibrium effect is reduced does not mean, however, that the phenomenon is irrelevant in these countries: firms spend a lot of resources in public relations to diffuse this threat.

Finally, we investigate the main mechanism through which the press had an effect. We find that, in roughly half of the cases, media pressure leads a regulator or a politician to intervene, while in the remaining half, it is the company itself that relents, realizing the reputational costs of continuing the battle. In sum, this evidence suggests that the primary mechanism through which media coverage has an effect is by increasing the reputational cost of misbehavior vis-à-vis a relevant audience (in this case Anglo-American investors).

This paper contributes to the literature on the real effects of media coverage. Previous work has looked at the impact of coverage on the voting behavior of citizens (George and Waldfogel (2004) and Della Vigna and Kaplan (2007)) as well as of representatives (Dyck, Moss, and Zingales, 2005). As Dyck and Zingales (2002, 2004), this paper looks at the impact of coverage on corporate governance. But rather than focusing on a cross-country correlation between newspaper circulation and various corporate governance outcomes, this paper focuses on a within-country setting where we are better able to identify the impact of the press. In this respect, our paper is similar to Miller’s (2006) and Dyck, Morse and Zingales (2007), as both include an examination of the role played by the media in bringing to light corporate frauds in the United States.

Our paper is also related to the growing literature on the determinants of possible media biases. Previous work has emphasized the biases generated by advertising pressure (Reuter and Zitzewitz, 2006), media ownership (Besley and Pratt (2006)), competition for audience (Baron (2005), Mullainathan and Shleifer (2005) and Gentzkow and Shapiro (2006)), and the quid-pro-quo between journalists and sources (Dyck and Zingales, 2004)). By contrast, this paper looks at the ability of financial institutions, with sufficient ‘skin in the game’, to influence whether a story makes its way to the international press.

Finally, our paper is also related to a large literature on shareholder activism. As nicely summarized by Gillan and Starks (2003) and Karpoff (2001), the bulk of this evidence has focused on pension and mutual funds and their attempt to discipline managers with traditional control mechanisms, such as incentive contracts (Almazan, Hartzell and Starks, 2005). As Kahan and Rock (2006) we study a new important player (hedge funds), but we focus on the interaction between this new player and an alternative mechanism: shaming in the press. In addition, our use of the Hermitage holdings as an instrument allows us to make further progress towards establishing a causal link between activism and outcomes. A limitation of our study, due to the illiquidity of the Russian market, is that we can only look at specific governance disputes rather than overall share performance.

The rest of the paper proceeds as follows.  In Section I we introduce a parsimonious theoretical framework for considering the impact of the media. We start by arguing that the media can matter, as they impact the reputation of the agents involved. In Section II we explain why we focus on the Russian market.  Section III describes our research design and data. Section IV studies the determinants of media coverage of major corporate governance violations and the impact that the Hermitage fund has on this coverage. Section V presents our main results on the effect of media coverage on the probability corporate governance violations are addressed. Section VI presents results when we instrument for coverage with the presence of the Hermitage fund. Section VII discusses the mechanisms through which media affect outcomes. Section VIII concludes.

I. What Role Can the Media Play in Corporate Governance?

A. The Role of the Media in Information Diffusion

The role of the media is to collect, select, certify, and repackage information. In doing so they dramatically reduce the cost economic agents face to become informed. When the Wall Street Journal reports a table with the quarterly performance of mutual funds, for instance, an investor does not have to spend time collecting all the pieces of information herself, but she can glance at them in a second, for the price of a dollar (plus the opportunity cost of the time spent reading). Furthermore, if there is a strong complementarity between news and entertainment, as it is often the case for hot or titillating topics, the media can make the cost of absorbing information negative by packaging news appropriately (Becker and Murphy (1993) and Dyck, Moss, and Zingales (2005)).

This dramatic reduction (if not elimination) of the cost of collecting information is very important since, in many situations, individual agents face a rational ignorance (Downs, 1957) paradox: the cost of becoming informed exceeds the benefit they can personally gain from that information. Hence, the media have the power to overcome the “rational ignorance” result (Dyck, Moss, and Zingales, 2005). By doing so, the media increase the number of people who learn about the behavior of other people, thereby increasing the effect of reputation. In the words of Justice Brandeis: “Publicity is justly commended as a remedy for social and industrial diseases. Sunlight is said to be the best of disinfectants; electric light the most efficient policemen.”[5]

A.1 Which Reputation?

Starting with Fama (1980), the finance literature has recognized the importance reputation plays in disciplining corporate managers. The early literature, Fama (1980) and Fama and Jensen (1986), emphasized managers’ reputation vis-à-vis potential employers, who will determine future jobs and wages. Even with recent declines in CEO tenure, CEOs do not hop from job to job frequently. Especially for CEOs of large companies, the probability of re-entering the labor market (and thus the importance of their reputation vis-à-vis future employers) is minimal. By contrast, career concerns might lead directors to act against the interest of shareholders. Since they are appointed by managers, they should care about their reputation vis-à-vis them.

More important, instead, is the role played by a manager’s (or a company’s) reputation vis-à-vis financial markets, as modeled by Diamond (1989, 1991) and Gomes (2000). To the extent a company needs to access financial market repeatedly, its reputation will affect the terms of future financing. Since these terms affect the profitability of a company and its ability to exploit future investment opportunities, they will be important even for self-interested managers.

Managers, however, seem to care not only about their reputation vis-à-vis the financial market, but also vis-à-vis society at large. As Dyck and Zingales (2002) argue, managers often bow to environmental pressures not because these are in the interest of shareholders, but because they do not want to face the private cost of being portrayed as “the bad guys”.

B. The Role of the Media in Corporate Governance

Consider a manager who has to decide whether to make a decision that might benefit her personally, but might hurt her reputation and trigger some legal punishment. A simple application of Becker’s (1968) model has that a manager will be dissuaded from such an action if and only if

(1) E (Private benefit) < E (Reputational cost) + E (Punishment)

=[pic]i learns about it + [pic]

where [pic]is the reputational cost of this action vis-à-vis group i, [pic] is the probability group i will receive the news about the manager’s action and will believe it , [pic] is the probability of enforcement, and P is the punishment in case of enforcement.

The media influence the right hand side of this equation in four ways. By publishing the news they can change[pic], i.e., the probability that a given action is known to a certain audience and so it carries a reputational cost. Of course, different media have different audiences, so each medium has a special impact on its own audience’s[pic]. If, for instance, a company is planning to raise new finance and it cares about the capital markets’ perception of its own action, it will be very sensitive to coverage in outlets that are read by the financial market community.

The media can affect the right hand side of equation (1) by increasing the reputational cost [pic] too. One way they achieve so is by spinning the news. When the business press chastised the lavish compensation of the former New York Stock Exchange chairman Richard Grasso, many of the same directors who approved the compensation changed their position and denounced it. What triggered this change was not only the diffusion of this information to a large audience (hence a change in [pic]), but also the negative characterization of Grasso’s pay package. This negative slant increased the reputational cost the directors faced and lead to their about face.

Another related way the press can change [pic]is by creating common knowledge. Many oligarches probably do not condemn a manager who dilutes outsiders, as long as he does it under the radar screen. In fact, they might even congratulate him for his cleverness, if he gets away with it. When the dilution becomes public knowledge, however, and it is criticized by the international press, the very same oligarchs feel obligated to condemn the violation as well (and shun the offender) to dissociate themselves from that type of behavior.

Third, the media can have an impact by changing the probability of enforcement[pic].[6] This impact arises through three channels. The first one is a simple extension of Fama’s model to politicians: they care about their future employers, i.e. the voters.[7] The second channel passes through the role of media in the battle between public interest and vested interests. A major reason why vested interests have so much power in political decisions is because of the “rational apathy” of voters (Downs, 1957). As Dyck, Moss, and Zingales (2005) argue, however, this rational apathy can be overturned by the media. By making political news entertaining, the media can overcome voters’ cost to become informed and, in so doing, reduce the power of vested interests. Once again, Richard Grasso’s very large compensation became entertaining news and made a much larger group of people aware of the potential conflict of interest intrinsic to the position of the NYSE chairman, who is in part a defender of the interests of the NYSE seat owners, and in part regulator. This new awareness substantially weakened the position of the NYSE lobbying effort to maintain its monopoly position.

The third channel arises because politicians do not care only about reputation vis-à-vis voters, but also their reputation (and their country’s reputation) vis-à-vis foreign countries. Russian President Putin, for example, also cares about his own reputation vis-à-vis the Western world and, in particular, the United States. Any news (especially if reported in the international press) that makes him appear weak or not in control of the situation undermines his credibility in international circles. Therefore, he will be more likely to take an action to address a problem if this problem is visible to the international community.

In sum, in the face of a corporate governance violation, a regulator who has to decide whether to intervene faces a trade-off very similar to equation (1). On the one hand, the private benefits of not enforcing are represented by the effort saved and the gratitude acquired from the company committing the violation. On the other hand, the regulator faces some reputational cost for being perceived as ineffective in her own job. In addition she faces the risk of a punishment, if her inaction violates a law and if this law is enforced. By diffusing the news of a corporate governance violation, the media expose the regulator’s lack of activity, increasing the personal cost of her inaction. The SEC, for example, started to ask the NYSE board about its compensation practices after the first news of Richard Grasso’s compensation was published in the Wall Street Journal. The publication of that news informed many people about the issue and created some awareness that the SEC was passive on this front. This awareness was sufficient to spurt the Commission into action.

Finally, the media can affect the right hand side of equation (1) by impacting the size of the penalty P. This is definitely true if a case goes to trial, because media can impact the mood of a jury. But it is also true whenever the enforcer has any discretion in the size of the punishment and she is influenced by her reputation vis-à-vis the public at large.

Note that all the terms on the right hand side of equation (1) are ex ante estimates. Hence, what will affect the decision to commit a corporate governance violation is a manager’s expectation of the likelihood the relevant players will learn about his decision and how harshly his decision will be judged. After the decision has been made, however, what determines the probability of reversal of this decision is the actual realization of those costs, which is greatly affected by the coverage in the media. Hence, the impact of the media is most visible (albeit not necessarily most important) in environments where managers underestimate ex ante the degree of intervention and influence of the media. As we will explain momentarily, this is exactly the case in Russia.

C. When are the Media Most Effective?

If we look at equation (1), the impact of media is greater when the media reach a larger number of relevant groups (i.e, groups with whom managers care to maintain a good reputation) and when the news reporting generates a greater increase in[pic]. In the language of the media, these two characteristics are diffusion and credibility. Ceteris paribus, the more people a medium reaches, the broader will be the reputational impact of its reports. To produce an increase in[pic], the news must come from a credible source, otherwise it is not believed. If we receive an e-mail coming from an unknown organization that accuses a famous professor of plagiarism, we are unlikely to believe it. If the same news were reported in the New York Times, we would be much more likely to believe it because the New York Times has developed a good reputation (some recent incidents notwithstanding).

Finally, the effectiveness of the media depends upon several characteristics of the surrounding environment. First, as discussed in Dyck and Zingales (2002) shaming works when society at large share the same set of values. French newspapers did not try to shame former President Mitterand for his long lasting extra marital affair because most French are willing to condone such behavior. As former president Clinton experienced first hand, this is very different from the shared norm in the United States. When it comes to corporate governance, shaming might reduce corporate governance violations if most people believe there is a social benefit in protecting shareholders. This is definitely the case in the United States and among the readers of the Wall Street Journal and the Financial Times, but it cannot be said for the average reader in Russia.

Second, the magnitude of the penalty that can be inflicted on a violator depends upon the frequency and the importance of business interactions. If a firm does not need external financing or external alliances, if it does not sell a consumer product and does not depend on the Government for its business, then it is isolated from any form of social enforcement. In all the other cases, its reputation is quite important for its profitability.

II. The Russian Case

A. Why Russia?

If enforcement is very effective and/or legal punishments are extremely severe, the manager’s expected cost of violating minority shareholders’ rights is such that managers will never do so. For this reason, it would be very difficult to try to identify any effect of the media in a country with very effective corporate governance rules.

The same is true, however, if the media have a long track record of imposing reputational penalties on managers who violate investors’ rights. The fear of these penalties will dissuade any manager from committing a violation. Ideally, thus, we would need a country that has very little or no legal enforcement and where, at the time a decision is made, the reputational costs of a decision are perceived to be very low.

Russia during the period late 1990s-early 2000s scores “well” on both dimensions. During this period the standard instruments to readdress corporate violations were either non- existent (derivative suits) or completely ineffective (for example, courts were easily corruptible, see Slinko, Yakolev, and Zhuravskaya, (2004)). As a result, corporate governance violations were very extreme, very common, and very visible. Hence, we can relatively easily assemble a sample of objectively bad governance decisions and follow them over time.

At the same time, Russian managers were just starting to learn how to deal with the press, in particular with the foreign press. Having been raised in an environment (Soviet Russia) where the media were reporting only what the party establishment wanted, they were unlikely to factor into their decisions the reputational cost the media could inflict.

No one illustrates this learning process better than Khodorkovsky, the former CEO of Yukos. At the beginning of his career, Khodorkovsky hated the press and kept it at distance. After one of his rare meetings with journalists, he declared: "It would be more pleasurable to meet a bunch of our unpaid workers in Siberia."[8] In August 1999, however, when the Bank of New York was accused of laundering money for several Russian companies, Yukos changed strategy, because it was concerned that "despite the absence of specific data, U.S. officials have taken the publications quite seriously -- a U.S. Congress hearing is scheduled for mid-September. A possible result of this hearing could be a decision to refuse Russia the financial aid of international financial institutions.”[9] Such attention spurred Yukos to hire a Western public relations agency and to start to fight back all the allegations in the media. Explaining Yukos decision to keep his company public and to pay more attention to investors and public relations, Khodorkovsky said: "First, there are not many very big private companies - and we want to be very big. Second, we need access to cheap capital and that means openness. Third, a big oil company has lots of workers, lots of ecological responsibilities. If it is opaque it is not going to be popular. Finally, there is the issue of nationalisation, which we can never ignore. A private company is a lot easier to nationalise than a public one."[10] Following this public relations campaign, Yukos started to be praised in Western media as a model of financial transparency and Khodorkovsky became the darling of the Western press. While this strategy was not sufficient in preventing Putin from seizing Yukos, it certainly made it more costly for him to do so.

As the Khodorkovsky quotes suggest, Russians care about their reputation vis-à-vis the international community for three reasons. First, they might want to access international markets (for financing, joint ventures, and even sale contracts). Second, as an insurance policy, both to protect the legitimacy of their holdings and to facilitate an asylum request in case they become persecuted in Russia. Third, for personal satisfaction. After becoming rich, executives in many developing countries seek broader acceptance in the international community by joining the World Economic Forum at Davos, seeking positions on the boards of trustees of prominent international institutions, and so on. Negative news reported in international media can also have the effect of ostracizing the executives from these desired social circles. While the Russian oligarch Vladimir Potanin was successful in his efforts to join the trustees of the Guggenheim Museum in April 2002, Oleg Deripaska was “disinvited” from participating in the Davos meeting, and was stripped of his designation as “one of the global leaders of tomorrow” following negative press coverage of civil lawsuits alleging bribery, money laundering, and worse (Financial Times 2001; Wagstyl 2002).[11]

Can weIidentify an Exogenous Shift in News Coverage?

In addition to the two factors mentioned above, Russia provides an excellent environment to identify the impact of the press on governance, because there exists a fund that consciously plays a media strategy: the Hermitage Fund.

Founded in 1996 as a generic hedge fund with a Russian focus, the Hermitage Fund changed its strategy and focus after the 1998 Russian crisis. In the words of its chairman: “Our basic approach is to thoroughly research and understand where the corporate malfeasance is taking place and then go to great pains to simplify the story so the average person can understand what is going on. One of the reasons that certain companies have gotten away with various violations in the past is that no one really understood what was happening because the stories were so complicated. We then share the stories with the press. By doing so, we want to inflict real consequences – business, reputational and financial” (Dyck, 2002).

In explaining why his strategy is successful in increasing coverage, he says:

“You have to understand that the press doesn’t know about the stories, have the ability to understand some of these complicated activities, or can’t afford to do research. We have a lot of money invested. We are affected. We can devote the resources to do what it takes to truly understand what is going on. Our goal is to frame the issue so that it is clear to everyone what has happened. We do talk to the Russian press, but our focus is on the international press” (Dyck, 2002, emphasis added).

Since the Hermitage fund focuses on generating coverage in those companies where it owns shares, the presence of the Hermitage fund among the shareholders of a company should represent an exogenous shift in news coverage, which can be used to identify the causal mechanism between news coverage and governance outcomes.

B. Does Hermitage Generate News?

For Hermitage to produce an exogenous shift in coverage, it must not only want to generate coverage, but also be successful in doing so. In talking with its chairman, we identified two mechanisms the fund has used: to be a helpful source and to become news.

C.1 Being a Helpful Source

One mechanism whereby Hermitage generates news is to conduct research and then present and document this information to a selected group of reporters. Becoming a source for information has the immediate effect of providing the specific news Hermitage wants to present and determining the time of the release of the news.

To illustrate the impact of Hermitage on news coverage, consider the coverage Gazprom, Russia’s largest company, received regarding some related-party transactions. There had been widespread concerns about Gazprom’s deals with related parties, but this became a focus of attention (and was finally addressed seriously) only when Hermitage provided crucial information to the press. In the words of Bill Browder, head of the Hermitage fund:

My head of research was able to buy the entire Moscow registration database from a hawker on a street corner. With the securities commission database, we knew the names of the companies that stole assets from Gazprom, and with the registration chamber data, we knew which individuals owned the companies. From that we were able to piece together exactly how much was stolen and by which members of management. … [We] decided to share our findings with the world by selectively releasing different examples of the graft to the major Western newspapers in Moscow. (Dyck (2002))

By October of 2000, Hermitage had put this information together in a 41 page PowerPoint presentation that laid out the story they wanted told, and presented the underlying information, including the sources. As Table A1 in the Appendix shows, there is a clear overlap between their information and resulting stories.

Not only did Browder present new information in his continuing campaigns, he also worked hard to time the presentation of information, and to ensure continued coverage of stories they cared about.

Originally, we would give one reporter the whole story. They would want to check every bit of it out, get the other side’s point of view, or ignore it, seeing this as too complicated and time consuming to pursue. Now we give a small piece of the story to a journalist and let them know that we’ll give it to someone else in three days if they don’t write anything. It seems that journalists are more concerned about losing the story to a competitor than almost anything else. (Dyck (2002))

Suggestive of the success of this strategy, we also see continued coverage of these allegations in the international news, as well as successful outcomes. Concrete steps were taken to limit the dilutions of Gazprom, including new requirements for board approval, new audits of these related party transactions, and the removal of the chief executive at the center of these allegations. Panel C of Table A1 provides a timeline of these outcomes. It also shows that this story, unlike so many other allegations of shareholder violation in Russia, did not die, but rather was repeated again and again over the next 6 months.

C.2 Becoming News

Another channel through which Hermitage generates news is by becoming news, filing a lawsuit. As Bill Browder argues,

We also go to courts. We’ve been involved in 32 lawsuits. And we win in terms of public attention regardless of the outcome, where we’ve lost 31 times. I think the proportion of number of words written in the press when a lawsuit is initiated to when it is dismissed is 50 to 1. The court of public opinion is much more effective than the Russian legal system and much fairer. (Dyck (2002))

The case of Sberbank illustrates this channel. At the end of 2000, the Sberbank board announced plans to go forward with a new share issue, which had the potential of diluting the ownership stakes of existing shareholders. It was hard for shareholders to fight against this decision using traditional methods, since there were no representatives of minority shareholders on the board. Hermitage, instead, chose to launch 12 different lawsuits against Sberbank and the Central Bank. Although the lawsuits were all dismissed, they generated a large amount of publicity, which came at a time when the Russian Duma was debating a new law on investor protection..[12]

Not only a lawsuit is news itself and induces newspapers to write about an issue, it also allows journalists to write about it without any fear of being sued for libel. If a journalist writes an article about dubious related party transaction, he might get sued by the company. But if he reports the same facts as the allegation made in a legal case, he incurs no risk. And this is a concern, since the first reaction of many Russian oligarchs to the bad Western press was to sue the journalists that wrote the articles in the court of London, a court more favorable to the plaintiff in libelous cases.

C. Selection vs. Causality

For marketing purposes, hedge funds have an obvious interest in self-promotion. To justify hefty management and performance fees to their own investors, hedge fund managers have to claim they have a strategy that adds value. For this reason, we have to be suspicious of Hermitage’s claim to be so successful in exposing corporate governance violations in the international press. To this purpose in section V.2 we will test whether it is indeed true that the presence of Hermitage as a shareholder leads to more coverage, after controlling for a series of companies’ characteristics.

Yet, finding such a correlation is not necessarily evidence of a causal link. An equally plausible explanation is that Hermitage buys into companies that are more visible or when it knows they will receive more attention from the press.

To minimize this concern, we follow two strategies. First, we construct a measure of newsworthiness and we insert it in the regression. Second, we choose to use the earliest Hermitage portfolio composition we have available, i.e. December 1998. This predates the major wave of corporate governance violations following the Russian crisis, and hence could hardly be thought as the result of an active strategy to pick more media sensitive companies. It also predates the period when Hermitage actively used the press as part of its strategy to increase returns in its portfolio.

III. Data Description

A. The Sample

Ideally, we would like a complete sample of corporate governance violations during a certain period. Fortunately, between 1998 and 2002 Troika Dialog, a prominent Russian investment bank, published the Bulletin on Corporate Governance Actions. This bulletin, usually averaging 8-10 pages in length, provided extensive coverage of all companies in Russian capital markets. It contained a one to two paragraph description written by specialists at Troika of corporate actions that came to their attention over the previous week. The Bulletin ceased publication in July 2002 as a result of the heavy lobbying done by Troika Dialog’s clients, who did not like the negative publicity generated by the Bulletin.

We read all sections in each issue of the Bulletin for the period December 4, 1998 to July 22, 2002. We focused primarily on the events reported in a sub-section titled “Reported/Potential Governance Violations”. But we also looked at other sub-sections, such as “New Share Issues” and “Split/Swap /Conversions”, and include an event whenever the paragraph description raised concerns that the proposed action would have a negative impact on the cashflows or voting rights of investors. This search resulted in a sample of 480 events.

We then refined the sample introducing a number of additional criteria. We eliminated all events that update earlier mentioned events, and dropped obvious minor events (e.g., minor delays in reporting) to arrive to 201 non-repeat potentially serious governance violations. We then read the English and Russian press to learn more about these events. We used this additional qualitative information to eliminate: (a) minor events, e.g., directors recommend one dividend level and AGM recommends another (32 observations); (b) events that we discovered upon further reading were not initiated by insiders but rather by government actions, e.g., the State blocks a shareholder from getting seats commensurate with his ownership stakes citing security concerns) (47 observations); (c) 24 additional observations where we were still left with uncertainty about the nature and severity of the potential violation. The remaining 98 events form the core sample for our study. Appendix B provides more details on this initial screening and the journal website provides a brief description of the alleged violations in these events and the date when this was first reported Table I defines the other variables used in the empirical analysis and reports their sources.

Reading all these allegations, we groups the different strategies used in seven categories. The first group includes all attempts to disenfranchise shareholders. This was done through a variety of mechanisms: threatening investors with imprisonment or worse if they don’t go along with wishes of controlling shareholders, not allowing shareholders to vote their shares, changing the venue at the last moment to make it impossible for shareholders to vote, and making decisions to sidestep shareholder approval for corporate transactions.

The other six categories include different methods to dilute the cashflow rights of minority shareholders. The first popular method consists in large share issues reserved to insiders at deep discount vis-à-vis the current market price. The second method consists of a share swap between companies and subsidiaries on terms that are viewed as hurting the interests of minority shareholders. The third method involves a reorganization of the firm and its subsidiaries that provides increased scope for self-dealing transactions. The fourth method involves debt holders using bankruptcy proceedings to reallocate assets to themselves, and in so doing diluting minority investors. The fifth method involves selling assets or business opportunities to companies closely affiliated with management. Finally, we include a sixth category, ‘other’, that captures other ways in which controlling shareholders dilute controlling shareholders.[13]

B. Performance Measures

To measure the impact of media pressure on the ability to contain or overturn the corporate decisions that violated minority shareholders’ rights, we look at the actual outcomes. After reading the Russian and the English language press over the subsequent year, we coded an outcome as 0 if the initial decision went through as planned and there were no governance changes in the firm linked to the (possible) furor surrounding this event. We coded an outcome as 2 if the proposed violation led to a significant response in the firm (11 instances). This includes reverting the initial decision, introducing significant changes in the terms of the transaction, or approving structural governance changes that makes further such actions unlikely (e.g. change in CEO, change in charter, change in number of independent board members, change in national law). We code an outcome as 1 if there is a partial redress of the shareholder concerns (17 cases). Overall, there was a positive outcome (partial or significant) in 29 % of the cases.

We thought to use the long-term performance of the stock price as a measure of the outcome. We discarded the idea for three reasons. First, the paucity of actively traded companies dramatically reduces the sample. Second, the timing of the possible reversal (stretching over months) make it difficult to identify out-performance, especially in an environment, such as the Russian one, characterized by high volatility in stock prices. Last but not least, if the market is so rational to anticipate the impact of media, the stock price performance will underestimate their effect.

C. News Measures

. The arguments exposed in section I suggest that we have to distinguish between three categories of the press: Russian media published in the Russian language, which have access to the Russian public but have limited credibility[14]; foreign-owned media in the Russian language, which have access to the Russian public but enjoy greater credibility; Anglo-American media, which have access to the international centers of economic and political power, where English is the lingua franca, and enjoy greater credibility.

For the Russian language press we focus on three large and/or prominent newspapers: Kommersant, Izvestia, and Vedemosti. We consider Vedemosti, to be the most credible as it is jointly published by the Wall Street Journal and by the Financial Times. For the English language news, we focus on the Financial Times and the Wall Street Journal as credible western news outlets. We measure the news coverage of the violation in a window surrounding the announcement of the event (t-1 to t+2 months). We use the period pre-dating our definition of announcement date to allow for the possibility that we may have misspecified the announcement date, particularly since the Bulletin is only a weekly publication. Our results are robust to just including t to t+2 months. For the most part we focus on the combined coverage in the English (Russian) press, although we also break down coverage by publication.

D. Reputation Measures

Since, in Russia, all legal remedies are very weak and the large shareholders are widely considered the villains rather than the monitors, an important source of restraint is given by the reputation these companies have vis-à-vis foreign investors. We try to capture the scope of such reputation concerns through two proxies: the percentage of equity owned by foreigners and the presence of the European Bank of Reconstruction and Development (EBRD) among the company’s lenders.

We obtained the EBRD investment information from the EBRD publication for Russia, “EBRD Investments: 1991- 2004” and indicate with a dummy if the EBRD had an investment prior to the commencement of the potential governance violation. We obtained the foreign ownership data from a variety of sources, starting with the official recording of the identities of all shareholders with more than 5 percent stake, collected by the Federal Commission on Security Market Disclosure project. In almost all cases, this was insufficient as many owners of record are shell companies with unclear ownership structure (e.g. Cyprus based companies) and/or are nominal owners (e.g. Citibank) without further information. We complemented this information with accounts in the business press (Russian and English) and in Troika Dialog “Bulletin on Corporate Governance Actions.” Recognizing the noise in this measure, we also assembled an alternative measure of foreign interest in the stock, which is a dummy variable that takes the value 1 if Troika Dialog collected and reported governance scores for the company sometime in the period from 2000 to end of 2002. It was time consuming and expensive to assemble such scores and Troika would only do so for companies with significant foreign investor interest. Below, we report results using the continuous measure of ownership, but the results are robust (and statistical significance strengthened), if we use this alternative indication of foreign interest. Table II presents the summary statistics of the variables used in the subsequent regression.[15]

E. Other controls

Other factors may affect the probability of reversal of a corporate governance violation. The nature of the violation itself, for example, can make it more or less reversible. To this purpose, as noted above, we have classified all the corporate governance violation into 7 categories.

Another factor that might influence the reversal of a corporate governance decision is the size of the potential loss inflicted to investors. To this purpose we create a variable measuring the maximum amount of potential dilution of shareholders’ value as a result of this decision. Specifically, for the companies where the potential violation is one of the six categories of dilution, we calculate what the loss would be, were the dilution to go through and this would lead to the worst outcome for minority shareholders.

For example, with new share issues at a very low price to company insiders, we know the number of shares issued and the price at which they are granted so we can straightforwardly calculate the percentage change in the minority shareholders’ claim over the firm’s cashflow. As an illustration, if the minority shareholders held 10 percent of the firm, and insiders issued 100% more shares at a zero price to insiders, then minority shareholders claims would now be for only 5 percent of the firm and would see a 50% dilution in their claim.

Sometimes this maximum percentage loss calculation involves more assumptions, as in the cases of bankruptcies or reorganizations. We made these assumptions based on the practices in Russia at the time, when solvent companies were pushed into bankruptcy and they valuable assets stripped at knock-down prices. On the bases of this fear, for instance, we set the maximum dilution for Chernorgorneft at 100 percent, since its main creditor had the possibility to force the viable oil producer into bankruptcy where it could transfer of all valuable assets to itself at knock-down price, leaving nothing but debts to the shareholders.

We found it much more difficult to identify the maximum possible loss for shareholders in those violations that we categorized as disenfranchisements, since these were usually the first stage of a longer process that could deprive shareholders of significant rights and returns. Here, we used our judgment to classify disenfranchisements into three categories, from most severe (coded as a’3’) to least severe (coded as a ‘1’).

Finally, the probability of reversal of a corporate governance violation can be affected by the visibility of a company. Visible companies are more likely to get the attention of the press regardless of any intervention of hedge fund managers and this attention may pressure these companies into reversing their decisions. As proxies for visibility we use a measure of size (the logarithm of the book value of fixed assets in 1999), a dummy for oil and gas industry (by far the most important and internationally attractive sector) and the “natural” newsworthiness (the logarithm of 1+ number of references to this company in the Wall Street Journal and the Financial Times during the 6 month period from January to end of June 1998, a time preceding our sample period and prior to the unique period surrounding the Russian default). [16] In controlling for these variables we are going to underestimate the effect of the press. Nevertheless, we include these variables to isolate the causal link.

IV. What Determines News in the International Press?

As we will see in the next section, the type of news coverage that influences governance outcomes is coverage in the international press. But what determines news in the international press? We address this question in Table III, where we analyze the determinants of press coverage. As a dependent variable, we use the natural logarithm of 1 plus the number of articles that appeared in the FT and WSJ around the event windows.

To explain a company’s coverage by Anglo-American newspapers, we insert a measure of size and a dummy for the oil and gas industry. A second potential driver of coverage in the international newspapers is the percentage of a company owned by foreigners and the presence of the EBRD among its lenders. Hence we control for both these variables.

Even controlling for all these variables, it is not obvious that we have captured the different level of “newsworthiness” of different companies. A candy factory called “Red October” might intrinsically make a better story than an oil company with an unpronounceable name such as Orenburgneft. Finally, to capture the different intrinsic newsworthiness of different corporate governance violation, we insert dummies for the seven different types of violations and a measure of the intensity of this violation.

As column 1 shows, the two main determinants of coverage are the size of the firm and the “natural” newsworthiness. Together they explain 47% of the variation in coverage. In column 2 we insert as an additional explanatory variable: the percentage of a company owned by the Hermitage fund at the end of the 1998 from the Hermitage consolidated financial statements[17] If our conjecture (and the claims of the Hermitage Fund chairman) is correct, the level of coverage should be higher when the Hermitage Fund owned a stake, because it has an incentive to intervene and prompt journalists to write stories. The Hermitage ownership variable enters in a positive way and it is statistically significant at the 5% level. One standard deviation increase in this variable almost doubles the expected level of coverage, raising the explanatory power of the regression to 53%.

In columns 3 and 4, we test whether the Hermitage presence has more effect on the WSJ or on the FT. The estimate of the impact of the Hermitage on the WSJ is twice as large as the one on the FT, and, given the paucity of data, this is not statistically significant for the FT while it is significant at the 1% level for the WSJ.

V. The Effect of Media on Outcomes

The second question we try to address is whether press coverage has any impact on the probability that a corporate governance violation is partially or completely addressed. The simplest way to test whether news coverage makes a difference is with a non-parametric test. Since the biggest difference is between companies whose violation is reported in the international media (17 of the cases) and violation that do not get reported, in Table IV, we split the sample along this dimension. 59% of the violations covered by international media are reverted against a mere 22% of the violation that are not covered. A Mann-Whitney test rejects that the two distributions are the same at the 1% level, (p=.0025). The same approach tells us that the Hermitage fund enjoys a much better record of reversal among its portfolio companies than average: 45% successes in the companies owned versus 24% success in the companies he did not own.

Of course, this approach does not factor other possible differences between the two samples. For this reason, Table V repeats the exercise in a standard regression format. In Table 1VA (column 1), we present our basic specification. The dependent variable is our measure of outcome (which can be either two, or one, or zero), hence we run an ordered logit. As control variables we insert two proxies for reputation (foreign ownership and EBRD as a creditor), two proxies for the visibility of the company (log of assets and dummy for the oil industry), 6 dummies for the type of violation and a measure of the severity of the violation.[18] For the disenfranchising action, our measure is an ordinal one. For the dilutive issues, it is the maximum potential loss due to dilution. Since the two measures are not homogenous, we insert them in the regression interacted with a dummy for the type of damage inflicted to the shareholders (dilutive or disenfranchising). Since we are able to construct this variable only for 94 companies, the sample is reduced to this number.[19]

Of all these variables, only one (the maximum loss due to dilution) is statistically significant. A one standard deviation increase in the maximum possible loss due to the dilutive decision raises the probability of seeing the corporate governance violation reverted by seven percentage points.

The lack of significance of all these variables is not too surprising. A rational manager will only commit violations he thinks he can get away with. So if the probability of having to revert the decision is significantly higher for certain types of violations, rational managers should commit fewer of those. Similarly, if the probability of having to revert the decision is significantly higher for certain companies, the managers of these companies should be more reluctant in abusing their shareholders’ rights.

To this basic specification, in column 2 of Table 1VA, we add a measure of foreign press coverage (number of articles published in the FT and WSJ in the period one month before to two months following the event).[20] Press coverage has a positive and statistically significant effect. One standard deviation increase in the number of articles published in foreign newspapers increases the probability of full redress (outcome = 2) by 14 percentage points. One more article increases this probability by 5 percentage points.

Given the mass of observations with zero coverage, in column 3, we re-run the same specification with a dummy variable for positive coverage instead of the actual number of articles. The result is very similar. In column 4, instead, we use the logarithm of 1 plus the number of articles, which seems to be a good compromise between the two previous specifications. The results are unchanged.[21]

Since we do not have a compelling theory of which news gets covered in the international press and which does not, it is possible that the effect of coverage is spurious. Companies that are more interesting to the media, for instance, can also be companies where the shareholders are better able to fight off managers’ violations. To account for this possibility in column 5 we add to the previous specification a measure of “natural newsworthiness”, which we measure as the natural logarithm of 1 plus the number of articles referring to this company in the Wall Street Journal and the Financial Times during the 6 month period from January to end of June 1998.[22] All the effect seems to be captured by the actual coverage, not by the newsworthiness.

The distinction between partial and full redress could be considered somewhat arbitrary, hence in the last column of Table 10 we collapse these two categories into one and we re-run the same regression as a logit. The results are virtually identical. The only difference is that we lose one observation since the dummy for violations that take the form of asset stripping perfectly predicts the outcome.

Finally, if the chosen window for the articles (a month before and two months after) includes the period in which the reversal takes place there might be a mechanical correlation between reversal and coverage, because reversal might lead to more coverage. When we checked for this possible overlap, however, we found only one case. In this instance one article included in data reported the reversal decision. If we eliminate this observation, however, all the results are the same.

In Table VB, we try to probe deeper into which articles are more important for a positive outcome. Column 1 inserts, as an explanatory variable, the coverage in Russian newspapers (log of 1 plus the combined number of articles in Kommersant, Izvestia, and Vedemosti.) The effect is negative but economically and statistically indistinguishable from zero.

Column 2 decomposes the effect of foreign press coverage between the Financial Times (FT) and the Wall Street Journal (WSJ). The coefficient of the WSJ is three times bigger than that of the FT and it is statistically different from zero (unlike the FT’s one). Even correcting for the higher mean and standard deviation of FT articles, the WSJ has more impact: one standard deviation increase in the number of WSJ articles increases the probability of a good outcome by 6 percentage points versus 3 percentage points for FT articles. One possible explanation for this difference is that much of the effect of the Financial Times is absorbed by the presence of the EBRD dummy. If we exclude that dummy, the coefficient of the FT does indeed increase, but remain 50% smaller than that of the Wall Street Journal.[23]

One of the Russian newspapers, Vedemosti, is a joint venture between the Financial Times and the Wall Street Journal, with both publishers appearing on the masthead below the Russian name. As such, it should enjoy a reputation similar to that of its owners. On the other hand, by being written in Russian language, it only circulates in Russia. This allows us to distinguish whether the difference in the impact of Anglo-American newspapers and Russian ones is due to differences in their credibility or in the audiences they have access to.

In Table VC (column I), we re-estimate the basic specification using, as a measure of press coverage, just Vedemosti (the log of one plus the number of articles published in Vedemosti). The coefficient is negative and not statistically significant. The same occurs in column II, where we also control for the coverage in other Russian newspapers. In fact, the coefficient on Vedemosti is lower than that of the other, less credible, Russian newspapers. Hence, keeping the audience constant, differences in credibility do not seem to have a major effect. Finally, in column III we insert also the coverage in Anglo-American newspapers. While the coefficient on Anglo-American newspapers is positive and significant, the coefficient on Vedemosti remains negative (albeit insignificant).

In sum, newspapers in Russian, even when credible, do not seem to play much of a role. Hence, we have to infer, consistent with the views of Hermitage’s chairman, that in Russia the main source of leverage on corporate managers is the access to an international audience.

VI. Addressing the Causality Problem

A. Instrumental Variables Estimation

There is a potential objection to the results presented in Table V: more severe violations are likely to trigger more news stories. If it is also true that more severe violations are more likely to be redressed, then our result may be spurious. In part, we have addressed this problem by controlling for the type of violation and the severity of the violation.

But to confront it directly, we now exploit the regressions on the determinants of news coverage described earlier and presented in Table III. Having identified some exogenous determinants of press coverage, we can now verify whether the effect of coverage on outcome is spurious by using this exogenous factor as an instrument. For comparability reasons, in column 1 of Table V, we reproduce the estimate of the basic specification using OLS. In column 2, we instrument foreign press coverage with the stake owned by the Hermitage fund in 1998. The IV coefficient remains positive and statistically significant, now at the 5% level. A one-standard deviation increase in coverage driven by the presence of Hermitage increases the probability that a corporate governance violation is reverted by 50%.

While we believe this instrument is exogenous, we address potential concerns in three ways. First, we appeal to the case study evidence in section II.C. In that section we show two mechanisms the Hermitage fund uses to generate coverage: producing reader-friendly reports and becoming news. The strong temporal association between the Hermitage actions and the published news and the striking similarities between the Hermitage reports and the published articles give support to the causality interpretation.

B. Hazard Function

Another way to address the causality problem is to focus on the specific timing of the three key events: the proposed governance violation, the news stories about it in the international press, and the possible reversal of the initial decision.[24] To this purpose, we build a duration model where the duration is the number of months between first public recognition of the possible violation and reversal (either partial or complete), if there is reversal. We first restrict ourselves to those companies where reversal takes place within six months, then we include in addition companies where reversal takes place between 6 months and one year, collapsing that time period into one observation.[25] We then use this duration as a dependent variable in a Cox proportional hazard model.

To explain this duration we include as our key explanatory variable the cumulative number of news stories in the Anglo American press, up to that point. As Table VII shows, the number of stories enters positively and significantly. Hence, more stories not only affect the probability of reversal but also affect the timing of reversal as well.

C. Other Implications

Yet one more way to examine the hypothesis of causality flowing from media to outcomes is to see whether there is evidence of learning about the power of media strategies in the time patterns of violations. For some corporate managers to be forced into changing their initial decisions, it must be the case that they were surprised by the intensity of the reaction to their corporate governance violation. If this learning is occurring in our sample, we should observe that the frequency of violations decreases over the sample period. Furthermore, if the Hermitage fund does indeed play a role in boosting the international reaction, the frequency of violation should drop more for companies that have the Hermitage fund among its shareholders.

To test these propositions we collected a second larger sample. This sample identifies all of the companies mentioned by name in the Troika Dialog Bulletin in all sections during our sample period. This sample of 493 companies includes firms that commit violations, firms against which violations are committed, and firms that are mentioned in passing as part of a discussion of ongoing events in the Russian market. This sample represents the universe of potential firms that Troika could have identified governance violations in.

Armed with this sample, we then identified, for each six month period, whether a violation had been committed, using as our sample of violations the 98 events mentioned above. Since our sample begins in November 1998, the first complete semester starts in January 1999. Figure 1 reports the temporal behavior of these frequencies for the sample of companies owned by Hermitage as of the end of 1998 and for the rest of the companies.

Both curves show a strong negative trend, as implied by the hypothesis that corporate managers learned about the power of the international press during the sample period. The frequency of violations among companies not owned by the Hermitage fund dropped from 4% to less than 1%. Among companies owned by the Hermitage, this drop was much more pronounced: from 15% to zero. Hence, companies owned by the Hermitage were much more likely to experience a corporate governance violation at the beginning of the sample period, but not so at the end. This graphic impression is also seen to be statistically significant if we run a linear probability model of corporate governance violation, with company fixed effects, a time trend and an interaction between the time trend and the presence of the Hermitage fund (regression not reported).

VII. How Does Press Coverage Lead to Better Outcomes?

The question we have not addressed yet is how press coverage succeeds in changing outcomes. In the Gazprom example, the coverage given to the corporate governance violation had several effects. First, the government officials on the Gazprom board felt compelled to side with minority shareholders and pass a motion that required board approval for any subsequent dilutions. Publication of this news also helped coordinate the actions of small institutional shareholders, who demanded an audit of these transactions. These stories inflamed the investment community, and helped Hermitage to convince other investors to sign their proxies to get the necessary 10% required to demand an independent audit of these and other transactions in December of 2000. The revelations also provided additional motivation for the government to change the CEO of Gazprom in May of 2001.

As this example illustrates, the mechanism through which media coverage affects outcomes is very complex. It is difficult to identify one single force. All the factors that played a role in the Gazprom case seem to have gained strength as a result of news coverage. Nevertheless, we can say that in this Gazprom case, the final difference was made by a political intervention.

Dyck (2002) provides further examples of how media pressure influences political actions, one notable example being the reversal of a proposed dilution by the Russian Securities regulator when Hermitage made the issue sufficiently high profile that he had the freedom to act. As Browder claims,

The reason he made this decision is that I was screaming bloody murder. He had a great scandal on his hands. Nobody had ever taken such a visible and outspoken position. I was shooting from the trenches, and this gave him cover to take his own steps. You have to remember that, as has become clearer since then, oligarchs owned the government and Vasiliev was worried about terrible things happening to him, professionally or even worse.[26] By not initiating but responding to an attack, he felt more empowered to act. He asked us on a number of occasions to raise specific points in the press because he couldn’t go on the offensive until something came out publicly. He was clear that he couldn’t be seen as initiating but responding. (Dyck (2002))

With this logic of trying to identify the primary channel which leads to redress (and caveat that all events are complex and open to multiple interpretations) in mind, in Table VIII, we try to group the positive outcomes according to the main force behind reversal. Roughly 36% of the cases reach a (at least partially) positive outcome as a result of the intervention of a regulator. What does press coverage have to do with the decision of a regulator to intervene? By overcoming Downs’ (1956) rational ignorance result, press coverage makes more people aware of the issues involved, increasing the regulator’s reputation costs of not acting (Dyck, Moss, and Zingales, (2005)).

Another 14% of the cases get resolved because of political intervention. If Russia were a typical democracy, the reasoning would be very similar. Politicians feel compelled to intervene on issues that are highly visible, because their political reputation is on the line. Being Russia (and being press coverage in a language not read by most of the voters), the reasoning might be different. The important factor here is the reputation vis-à-vis foreign (and in particular Anglo-American) investors and policy makers. Even Putin has some concerns about the way he is perceived in the West.

In addition, coverage of certain events by the foreign press might provide political coverage for a Government intervention motivated by other reasons. In the UES case, for instance, Putin took advantage of the negative coverage that his political rival Anatoli Chubais, chairman of UES, was receiving in the Western press and intervened to reduce his power. Had foreign media not attacked Chubais, Putin might have been more reluctant to intervene for fear of the repercussions his actions might have on his reputation in the West.

In another 29% of the cases, a positive resolution is due to effect of press coverage on the pre-existing opposition. For example, in the Kamaz case, the EBRD was fighting the share dilution approved by the company. Press coverage strengthened the EBRD case because it increased the awareness of investors as to the behavior of Kamaz and in so doing increased the reputation cost of misbehavior.

In the remaining 21% of the cases, it looks like the company voluntarily changed its course of actions. In these cases it is more difficult to establish what role did press coverage play.

In sum, it looks like the primary mechanism through which media coverage has an effect is by increasing the reputation cost of misbehavior vis-à-vis a relevant audience (in this case Anglo-American investors). Obviously, the success of this strategy is highly dependent on the importance the key actors attribute to their reputation vis-à-vis this constituency.

VIII. Conclusions and Implications

Our paper aimed at establishing four facts. First, news coverage is driven not only by the intrinsic appeal of each piece of news, but also by the lobbying effort exerted by those with an interest in the news being published. Second, media coverage is not just a mirror of reality, but it can have important effects on reality itself, in particular on corporate governance. Third, media coverage is effective only when a behavior violates norms widely accepted in society. Fourth, the effect of media can be economically large. We show that in Russia, during the period 1998-2002, all these four statements are true.

That news, even news in the most reputable newspapers, can be influenced by the lobbying of different economic interests may not come as a surprise. But, as far as we know, it has not been documented empirically before. What might be more surprising is the magnitude. We show that one single fund, with a 10 percent stake and a very limited investment of resources, can double the amount of coverage an event receives.

What is special about this time period in Russia is that firms were relatively unsophisticated in counterbalancing the activism of the Hermitage fund. This situation, however, is not unique. There are many firms, both in developing and developed countries, that do not want to interact with the media, [27] do not know how to do it, or are overpowered in their ability to do so by the opposite side. This latter situation is increasingly more common with the emergence of activist hedge funds. Our estimates are likely to apply to all those cases. And while the magnitude of our estimates is unlikely to be true when firms fight back in the public relations war, it does illustrate the size of the stakes involved. If firms spend lavishly on public relations, it is because activists can have an effect as large as the one documented in our paper.

That newspapers are not only a mirror of reality but they can shape it is also widely believed. The unique contribution of our paper, however, is to isolate a situation in which we can identify this link, through the use of instrumental variables. We see our effort as a first step in this direction. While all the robustness tests and the case study evidence we collect corroborate the validity of our instrument, it is impossible to infer causality from just one study. More research is needed.

Our natural experiment allows us not only to identify when media pressure has an effect but also when it does not. We show that Russian media, even when credible, do not seem to have an impact, while foreign media do. Our conjecture is that media shaming might be effective in reducing corporate governance violation only if most people believe it is socially valuable to protect minority shareholders. When such a belief is not present in society (as was not present among Russians), shaming does not work.

In sum, our analysis suggests that in most developing countries, with little or no credible media outlet and an unsophisticated public opinion, domestic media are not very powerful in pressuring companies to behave. Interestingly, however, even in these countries, foreign media can have an impact, as long as companies need credibility to raise capital or establish joint ventures abroad. Our findings are of particular importance for countries like China with unformed legal systems.

A direct implication of this distinction is that in developing countries, the better governed firms will be those that need to access the international capital markets. By contrast, in developed markets the better governed firms will be all those that need to access capital markets, regardless of whether domestic or international. This hypothesis is consistent with Dyck and Zingales (2004)’s finding that on average companies are better governed in countries with more diffused (and hence more credible) press.

Last and not least, our paper shows how strong the effect of media reporting on corporate governance can be. One more article in the FT or the WSJ increases the probability of reverting a corporate governance violation by five percentage points. We estimate that the value of an extra article published in the Wall Street Journal or the Financial Times is $ 40.4 million (3.3 million).

There is no reason to believe that this effect is unique to Russia. In fact, given the poor institutional environment and the lack of pro market values among Russians, one would expect this to be an underestimate of the effect media can have on corporate governance in countries like the United States, where media are much more credible and society share a pro shareholder set of values.

References

Almazan , Andres, Jay C. Hartzell and Laura T. Starks, 2005, Active Institutional Shareholders and Costs of Monitoring: Evidence from Executive Compensation, Financial Management 34, 5-34.

Baron, David P., 2005, Competing for the Public Through the News Media, Journal of Economics & Management Strategy 14, 339-376.

Baron, David P., 1996. Business and its Environment (Prentice Hall: Upper Saddle River, New Jersey.).

Becker, Gary S., 1968, Crime and Punishment: An Economic Approach, Journal of Political Economy 76, 169-217.

Becker, Gary S. and Kevin M, Murphy, 1993, A Simple Theory of Advertising as a Good or Bad, The Quarterly Journal of Economics 108, 941-964.

Besley, Timothy, and Andrea Prat, 2006, Handcuffs for the Grabbing Hand? The Role of the Media in Political Accountability, American Economic Review 96, 720-736.

Bhattacharya, Utpal, and Hazem Daouk, 2002, The World Price of Insider Trading, Journal of Finance 57, 75-108.

Bhattacharya, Utpal, Hazem Daouk, Brian Jorgenson and Carl-Heinrich Kehr, 2000, When an Event is Not an Event: The Curious Case of an Emerging Market, Journal of Financial Economics 55, 69-101.

Della Vigna, Steffano and Ethan Kaplan, 2007, The Fox News Effect: Media Bias and Voting, Quarterly Journal of Economics 122, forthcoming.

Diamond, Douglas W., 1989, Reputation Acquisition in Debt Markets, Journal of

Political Economy 97, 828-862.

Downs, Anthony, 1957. An economic theory of democracy (Harper & Brothers, New York, NY.).

Dyck, Alexander, 2002, The Hermitage Fund: Media and Corporate Governance in Russia, HBS case # N2-703-010.

Dyck, Alexander, and Luigi Zingales, 2002, The Corporate Governance Role of the

Media, in Roumeen Islam, ed.: The right to tell: The role of the Media in Development (The World Bank, Washington, DC.).

Dyck, Alexander, and Luigi Zingales, 2003, The Media and Asset Prices, Chicago GBS Working paper.

Dyck, Alexander, and Luigi Zingales, 2004, Private Benefits of Control: An International

Comparison, Journal of Finance 59, 537-600.

Dyck, Alexander, David Moss, and Luigi Zingales, 2005, Media vs. special interests, Chicago GBS Working paper.

Dyck, Alexander, Adair Morse, and Luigi Zingales, 2007, Who Blows the Whistle on

Corporate Fraud?, NBER Working Paper 12882.

Fama, Eugene F., 1980, Agency Problems and the Theory of the Firm, Journal of Political Economy 88, 288-307.

Fama, Eugene F., and Michael Jensen, 1983, Separation of Ownership and Control, Journal of Law & Economics 26, 301-25.

Gentzkow, Matthew A., and Jesse M. Shapiro, 2006, Media Bias and Reputation, Journal of Political Economy 114, 280-316.

George, Lisa and Joel Waldfogel, 2003, Who Affects Whom in Daily Newspaper Markets?, Journal of Political Economy 11, 765-785.

Gillan, Stuart and Laura Starks, 2003, Corporate Governance, Corporate Ownership and the Role of Institutional Investors: A Global Perspective, Journal of Applied Finance 13, 4-22.

Gomes, Armando, 2000, Going Public without Governance: Managerial Reputation Effects, Journal of Finance 55, 615-646.

Guriev, Sergei, and Andrei Rachinsky, 2005, The Role of Oligarchs in Russian Capitalism, Journal of Economic Perspectives 19, 131-150.

Hamilton, James, 2004. All the news fit to sell (Princeton University Press, Princeton,

NJ.).

Kahan, Marcel, and Edward Rock, 2006, Hedge Funds in Corporate Governance and Corporate Control, University of Pennsylvania, Inst for Law & Econ Research Paper 06-16.

Karpoff, Jonathan M., 2001, The Impact of Shareholder Activism on Target Companies: A Survey of Empirical Findings, University of Washington - Business School Working Paper.

LaPorta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert Vishny, 1998, Law and Finance, Journal of Political Economy 106, 1113-1155.

Miller, Greg, 2006, The Press as a Watchdog for Accounting Fraud, Journal of Accounting Research 44, 1001-1033.

Mullinaithan, Sendhil, and Andrei Shleifer, 2005, The Market for News, American Economic Review 95, 1031-1053.

Reuter, Jonathan, and Edward Zitzewitz, 2006 , Do Ads Influence Editors? Advertising and Bias in the Financial Media, The Quarterly Journal of Economics 121, 197-227

Shleifer, Andrei, 2005. A Normal Country: Russia After Communism (Harvard University Press, Cambridge, MA.).

Slinko, Irina, Evgeny Yakolev and Ekaterina V. Zhuravskaya, 2004, Laws for Sale:

Evidence from Russia, American Law and Economics Review 7, 284-318.

Wagstyl, Stefan, 2002, The Road to Recognition, Financial Times April 6, 1.

Appendix A: How Does Hermitage Generate News?

Panel A and B present slides from a 41 slide Hermitage Powerpoint presentation produced in early 2000 that detailed a series of self-dealing actions by company management involving Stroytransgaz, Itera, Rospan, Zapsibgazprom and other entities. This information was provided to journalists at the Financial Times and Wall Street Journal. Panel C illustrates newspaper coverage in the Financial Times and the Wall Street Journal that picks up on this news.

Panel A - Generating Initial News – Example 1

Hermitage Slide

[pic]

News Reporting Based on Slide

Financial Times reporting, October 25, 2000

“The FT has learned of one case in which Gazprom has been awarding large contracts to a company which is majority owned by the relatives of Gazprom’s past and present management. Stroytransgaz documents from 1999 show that over 50 percent of the pipeline construction company is owned by people close to Gazprom’s senior management. These include 6 per cent each by Vitaly and Andrei Chernomyrdin, the sons of the former prime minister, and former head of Gazprom Viktor Chernomyrdin, and a further 6.4 per cent by Tatyana Dedikova, the daughter of Gazprom’s chief executive Rem Vyakhirev. A further 20 per cent is held by Arngolt Bekker, Stroytrransgaz’s chief executive and a Gazprom director, and three of his relatives hold another 12.3 percent between them.…”

Financial Times reporting, October 28, 2000

“The Federal Securities Commission, the market regulator, yesterday confirmed it had asked for an explanation from Gazprom following the publication of an article in Wednesday’s Financial Times on relations between the group and Stroytransgaz, a construction company in which present and past senior managers of Gazprom and their relatives are significant shareholders.

Wall Street Journal, October 30, 2000

“Revelations that top Gazprom managers and their family members own a big chunk of Gazprom main pipeline-building contractor, OAO Stroytransgas, also caused concern among investors last week. Gazprom has awarded many lucrative contracts to Stroytransgas…”

Panel B - Generating Initial News – Example 2

Hermitage Slides

[pic][pic]

News Reporting Based on Slides

Wall Street Journal reporting, October 24, 2000

Investors holding about 20% of Gazprom complain that Itera has been allowed to gobble up valuable Gazprom assets on the cheap. Take Rospan, a joint venture holding licenses to two fields with reserves of 230 billion cubic meters of natural gas and 80 million metric tons of gas condensate. In 1998, Gazprom decide it couldn’t afford to invest in Rospan production and sold its 51% stake to two shell companies founded and owned by Itera: ZAO STI-Sigma and OOO Lanka-Promkomptekt. Despite Rospan’s rich gas reserves, MR. Vyakhirev ordered the stake sold at its nominal price, or par value, of 4,258 rubles -- $284 at the time. Gazprom’s minority shareholders value the lost gas and gas condensate reserves at $345 million.

[pic]

Wall Street Journal reporting, October 30, 2000

“The board briefly discussed the erosion of Gazprom’s stake in Zapsibgazprom, A Siberian firm whose 700 billion cubic meters of gas reserves are valued at about $600 million. Gazprom didn’t take up its rights in a recent share issue, allowing its stake to fall to 37% from 51% according to securities commission documents. Two shell companies of unknown ownership acquired 21% of the company during the issue. The price of the new stock issued puts the company’s market capitalization at just $12 million.”

Panel C – Continued Reporting of Allegations

This panel plots the number of articles published in the Wall Street Journal and in the Financial Times following the first announcement of a corporate governance violation at Gazprom.

[pic]

Appendix B: Sample Construction

This table shows how we arrived at our final sample based on an initial sample of 480 potential governance violations from the “Bulletin on Corporate Governance Actions” published weekly by the Russian investment bank Troika Dialog, for the period December 4, 1998 to July 22, 2002 (Bulletin #23-#142). This sample was based primarily on all the events reported in a sub-section titled “Reported/Potential Governance Violations”, “New Share Issues”, and “Split/Swap/Conversions.” The table identifies the criteria for excluding observations, the number of observations removed from the sample, and, where relevant, examples of excluded companies.

|Reasons for excluding observation |Examples |Number of |

| | |observations |

| | |480 |

|Event is an update of an earlier | |176 |

|mentioned event | | |

|Only reason for inclusion is that firm | | |

|has an ADR in association with the money | | |

|laundering investigation in the United | | |

|States | | |

|Minor delay in financial reporting | |89 |

|Identified company is the one committing |Yukos, March 1999 - “Proposed share swap with subsidiaries causing|14 |

|the governance violation. |rift among shareholders” | |

| |Sibneft, October 2001 - “Reverse split scheme comes under | |

| |criticism from Noyabrskneftegaz small shareholders” | |

|Subtotal of non-repeat, potentially serious governance violations = 201 |

|Uncertainty - Description of event |Aphipsky refinery, Sep 1999 – auction of this asset owned by |24 |

|presents conflicting or unclear |Rosneft subsidiary cancelled at last moment (but unclear why or | |

|information that could not be clarified |what is going on here) | |

|through additional investigation |Kauchuk, Dec 1998 – Company in bankruptcy but some assets slated | |

| |for auction are seized by creditors | |

|Minor event – Description of event |Ammonfos, April 2000 – Directors recommend one dividend level and |32 |

|suggests minor event |AGM recommends another in apparent violation of law. | |

| |Aviastar, April 1999 – Government report suggests company should | |

| |issue shares to dilute and limit foreign ownership, but rejected | |

| |by company. | |

| |Lensvyaz, Nov 1999, Event highlighted is that finally pay | |

| |dividends that promised earlier. | |

|Not valid event – Description suggests |Kaztransgaz, March 2000 – appears violation of contract with |47 |

|event not a governance violation |western JV partner, (with compensation) | |

|initiated by insiders but rather |Krasnoye Sormovo, June 1999- Repeated efforts by controlling | |

|government action or action consistent |shareholder to get seats on board commensurate with stakes blocked| |

|with contract |by state, that insists on keeping board members given to company | |

| |at time of privatization to protect state secrets over submarine. | |

|Observations included in data set = 98 |

Figure 1: Evolution of the Frequency of Corporate Governance Violations in Russian Firms

This figure plots the frequency of governance violations in two sub-samples of Russian firms. The frequency in each semester is determined by dividing the total number of companies classified as corporate governance violators during that semester using companies identified by the Russian investment bank Troika Dialog in their weekly “Bulletin on Corporate Governance Actions,” (December 1998- June 2002) by the total number of companies covered by Troika Dialog during the same period. The companies committing violations are those described in Table 2. The first sub-sample consists of the companies where the Hermitage fund had a stake in at the beginning of the sample period (based on their portfolio as of end of 1998), the second one by the rest of the Russian companies followed by Troika Dialog where Hermitage had no stake at the end of 1998.

[pic]

Source: Calculations by based on data from Troika Dialog, Hermitage Capital.

Table I: Variable Definitions and Sources

|Variable |Definition |Source |

|Type of governance violation |Based on a reading of the event reported in Troika |Troika Dialog’s “Bulletin on Corporate |

| |Dialog’s “Bulletin on Corporate Governance Actions” |Governance Actions” + Russian and English |

| |and news stories about the event in Russian and |news sources in Factiva. |

| |English news sources, we code 7 mutually exclusive | |

| |categories of alleged violation, which include | |

| |disenfranchisement and six types of dilution (Share | |

| |Issuance, Share Swap, Reorganization, Bankruptcy, | |

| |Asset Stripping and Other). | |

|Maximum loss due to dilution |The maximum potential loss variable assumes the |Federal Comission on Security Market |

| |proposed action went through and the worst fears were |Disclosure project, and Troika Dialog |

| |realized for minority shareholders. |“Bulletin on Corporate Governance Actions.” |

|Maximum loss due to |When the action is a disenfranchisement, we use a |Same as above |

|disenfranchisement |three point scale from lowest (1) to highest (3) | |

| |severity of the potential loss. | |

|Outcome of potential governance |For each alleged violation we code the outcome equal |Troika Dialog’s “Bulletin on Corporate |

|violation |to 0 if it was not redressed at all, 1 if partially |Governance Actions” + Russian and English |

| |redressed, and 2 if substantially redressed. |news sources in Factiva. |

|News coverage of alleged |The number of articles in English newspapers (WSJ and |Factiva, ISI Emerging Markets |

|governance violation |FT) and in Russian papers (Kommersant, Izvestia and | |

| |Vedemosti) is based on a count of the number of | |

| |articles mentioning a company in the specified | |

| |newspaper during the period from t-1 month to t+2 | |

| |months after the date of the alleged violation. We | |

| |read the complete text of all articles with the | |

| |company name, and retained only those articles that | |

| |made a reference to the alleged violation. | |

|Newsworthiness |Number of articles mentioning a company in the in |Factiva, ISI Emerging Markets. |

| |English newspapers (WSJ and FT) in a period prior to | |

| |the violation and before the Russian currency crises, | |

| |which we define as January-July 1998. | |

|Foreign ownership stake |Proportion of stock held by foreign investors. |Official recording of the identities of all |

| | |shareholders with more than 5 percent stake, |

| | |collected by Federal Commission on Security |

| | |Market Disclosure project complemented with |

| | |accounts in the business press (Russian and |

| | |English) and in Troika Dialog “Bulletin on |

| | |Corporate Governance Actions,” |

|EBRD dummy |Dummy variable equal to 1 if the EBRD has provided |EBRD Investments: 1991-2004, which lists |

| |loan financing to the company prior to committing the |loans by company and date. |

| |infraction. | |

|Log of assets |Log of assets is the log of book value of fixed assets|Federal Commission on Security Market |

| |in 1999. |Disclosure Project |

|Hermitage stake |The table indicates with a dummy variable whether |Hermitage Fund Consolidated Financial |

| |Hermitage Capital had a stake in the company based on |Statements, 1998. |

| |their reported portfolio composition at the end of | |

| |1998, the earliest available date for the Hermitage | |

| |fund. | |

Table II: Summary Statistics

This table reports summary statistics for variables included in tables 8-10. Foreign ownership is the proportion of stock held by foreign investors. EBRD is a dummy variable equal to 1 if the EBRD has provided loan financing to the company prior to committing the infraction. Log of assets is the log of fixed assets for 1999 in thousands of rubles (exchange rate=22.3 rubles per $). The oil industry dummy identifies firms in the oil and gas industry. The number of articles is based on a count in the period from t-1 month to t+2 months after the proposed violation. The measures of natural newsworthiness are based on the same publications in January-July 1998. We include 7 mutually exclusive categories of alleged violation, which include disenfranchisement and six types of dilution (Share Issuance, Share Swap, Reorganization, Bankruptcy, Asset Stripping and Other). The maximum potential loss variable assumes the proposed actions went through and the worst fears were realized for minority shareholders. When the violation is a disenfranchisement, we use a three point scale from lowest (1) to highest (3) severity of the potential loss.

|Outcome |Mean |Median |Standard |Minimum |Maximum |Number of |

| | | |Deviation | | |obser-vations |

|Foreign ownership |0.13 |.1 |0.16 |0 |.84 |98 |

|Dummy for EBRD stake |0.12 |0 |0.33 |0 |1 |98 |

|Log assets |14.88 |14.87 |1.99 |10.07 |20.9 |98 |

|Oil Industry dummy |0.01 |0 |0.02 |0 |1 |98 |

|# articles in FT & WSJ |0.71 |0 |2.18 |0 |16 |98 |

|Log of (1+ # articles in FT+WSJ) |0.25 |0 |0.6 |0 |2.83 |98 |

|Natural newsworthiness (log of (1+ # articles |0.5 |0 |1.1 |0 |4.71 |98 |

|in FT+WSJ) January-July 1998) | | | | | | |

|Log of (1+ # articles in FT) |0.2 |0 |0.5 |0 |2.48 |98 |

|Log of (1+ # articles in WSJ) |0.12 |0 |0.36 |0 |1.79 |98 |

|Log of (1+ # articles in Russian press) |0.77 |0.69 |0.77 |0 |3.56 |98 |

|Log of (1+ # articles in Vedomosti) |0.43 |0 |0.56 |0 |2.64 |98 |

|Type of Dilutions | | | | | | |

|Share issuance |0.29 |0 |0.45 |0 |1 |98 |

|Share swap |0.11 |0 |0.32 |0 |1 |98 |

|Bankruptcy |0.1 |0 |0.3 |0 |1 |98 |

|Reorganization |0.06 |0 |0.24 |0 |1 |98 |

|Asset stripping |0.02 |0 |0.14 |0 |1 |98 |

|Other forms of dilutions |0.1 |0 |0.3 |0 |1 |98 |

|Disenfranchisement |0.32 |0 |0.47 |0 |1 |98 |

|Maximum Loss in Dilution |0.57 |50 |0.33 |5 |100 |63 |

|Maximum loss in Disenfranchisement |2.06 |2.0 |0.77 |1 |3 |31 |

Table III: The Determinants of Press Coverage

The dependent variables are different measures of news coverage. In the first two columns it is the log of 1 plus the number of articles in the Financial Times and the Wall Street Journal in the period from t-1 month to t+2 months around the date of the alleged violation. In the third column the dependent variable is one plus the number of articles in the Financial Times during the same period. In the last column it is one plus the number of articles in the Wall Street Journal during the same period. Newsworthiness is the log of the number of references to a company in the WSJ and FT in the 6 month period from January to end of June 1998. Hermitage is the percentage of Hermitage portfolio invested in the company as of end of 1998. All the other control variables are defined in Table 3. When the violation takes the form of a disenfranchisement, we use a three point scale from lowest (1) to highest (3) severity of the potential loss. All the estimates are obtained by OLS. Huber/ White robust standard errors are reported in brackets, * means significant at 10%, ** 5% and *** 1%.

| |Log of (1+ # articles in FT and|Log of (1+ # |Log of (1+ # |

| |WSJ) |articles in FT) |articles in WSJ) |

| |I |II |III |IV |

|Natural "newsworthiness" |0.248*** |0.128 |0.163* |(0.020) |

| |(0.080) |(0.092) |(0.082) |(0.052) |

|Percentage of Hermitage assets | |11.340** |6.295 |11.285*** |

|invested in the company (1998) | |(5.486) |(4.674) |(3.445) |

|Foreign ownership (%) |0.555 |0.572* |0.344 |0.309* |

|  |(0.341) |(0.301) |(0.252) |(0.185) |

|EBRD as an investor dummy |0.030 |0.126 |0.147 |-0.025 |

| |(0.182) |(0.175) |(0.148) |(0.113) |

|Log of assets |0.087*** |0.049* |0.044* |0.014 |

|  |(0.032) |(0.026) |(0.022) |(0.017) |

|Dummy for oil industry |0.189 |0.222* |0.175* |0.106 |

| |(0.126) |(0.123) |(0.098) |(0.081) |

|Controls for nature of violation |

|Share issuance dummy |-0.059 |-0.029 |-0.02 |-0.006 |

|  |(0.179) |(0.170) |(0.147) |(0.101) |

|Share swap dummy |0.111 |0.116 |0.100 |0.124 |

| |(0.339) |(0.344) |(0.292) |(0.192) |

|Bankruptcy dummy |-0.028 |-0.006 |0.023 |0.01 |

|  |(0.258) |(0.252) |(0.215) |(0.137) |

|Reorganization dummy |-0.171 |-0.213 |-0.147 |-0.061 |

| |(0.244) |(0.221) |(0.190) |(0.123) |

|Asset stripping dummy |-0.274 |-0.628*** |-0.326* |-0.514*** |

|  |(0.261) |(0.208) |(0.180) |(0.130) |

|Other form of dilution dummy |-0.323 |-0.403 |-0.262 |-0.176 |

| |(0.340) |(0.318) |(0.272) |(0.188) |

|Maximum loss due to dilution |0.001 |0.001 |0.001 |0.000 |

|  |(0.002) |(0.002) |(0.002) |(0.001) |

|Maximum loss due to disenfranchisement |-0.07 |-0.068 |-0.049 |-0.039 |

| |(0.069) |(0.061) |(0.055) |(0.031) |

|Observations |94 |94 |94 |94 |

|R-squared |0.468 |0.526 |0.522 |0.482 |

Table IV: International Press Coverage and Outcomes

This table summarizes two key features of our sample of 98 observations: whether there was coverage in the international press; and, the type of outcome. The variable ‘coverage in the international press’ takes on the value 1 if there was any coverage in the Wall Street Journal or Financial Times in the period from t-1 month to t+2 months around the date of the alleged violation and 0 otherwise. In this table we club together outcomes coded as 1 (partial redress) and 2 (full redress) as positive outcome.

| |Positive Outome:|Outcome: Not |Total Number of |Percentage of |

| |Paritally or |blocked |Observations |Observations |

| |Fully Blocked | | |where Positive |

| | | | |Outcome |

|Media coverage in the International Press |10 |7 |17 |0.59 |

|No media coverage in the International Press |18 |63 |81 |0.22 |

|Total |28 |70 |98 |0.29 |

| |9 |11 |20 |0.45 |

|Companies where Hermitage fund has a stake | | | | |

| |

|Two-sample Wilcoxon rank-sum (Mann-Whitney) test |

| |

|Null hypothesis outcome (coverage==0) = outcome (coverage==1) |

|z = -3.021, Prob > |z| = 0.0025 |

Table V: The Effect of Press Coverage on Outcomes

The dependent variable is the outcome of the proposed governance violation. This outcome variable is equal to 0 if the potential governance violation was not redressed at all, 1 if partially redressed, and 2 if substantially redressed. The estimations are obtained using an ordered logit. In column VI of panel A we club outcome 1 and 2 and we run a simple logit. The number of articles is based on a count in the period from t-1 month to t+2 months around the date of the alleged violation. News coverage in English is a dummy variable that takes the value 1 if there was any news coverage and 0 otherwise. Newsworthiness is measured as the log of one plus the number of references to a company in the WSJ and FT in the 6 month period from January to end of June 1998. All the other control variables are defined in Table 3. Huber/ White robust standard errors are reported in brackets, * means significant at 10%, ** 5% and *** 1%.

Panel VA:

| |I |II |III |IV |V |VI |

|Total English articles | |0.522*** | | | | |

|t-1 to t+2 months | | (0.153) | | | | |

|News coverage in English | | |1.725*** | | | |

|Dummy (t-1 to t+2 months) | | |(0.633) | | | |

|Log of (1+ # articles in | | | |1.644*** |1.569*** |1.956*** |

|FT or WSJ) | | | |(0.425) |(0.505) |(0.725) |

|Natural | | | | |0.074 |0.056 |

|"newsworthiness" | | | | |(0.250) |(0.337) |

|Foreign ownership (%) |0.021 |-0.358 |-1.034 |-0.886 |-0.829 |-0.87 |

|  |(1.906) |(1.971) |(2.065) |(1.982) |(1.974) |(1.737) |

|EBRD as an investor dummy |0.876 |1.198* |0.815 |1.024 |1.059 |1.994** |

| |(0.658) |(0.669) |(0.668) |(0.668) |(0.691) |(0.876) |

|Log of assets |0.134 |-0.15 |-0.042 |-0.168 |-0.18 |-0.191 |

|  |(0.158) |(0.153) |(0.139) |(0.151) |(0.160) |(0.175) |

|Dummy for oil industry |-0.283 |-0.677 |-0.675 |-0.866 |-0.862 |-0.708 |

| |(0.739) |(0.806) |(0.767) |(0.812) |(0.804) |(0.730) |

|Controls for nature of violation |

|Share issuance dummy |0.003 |-0.004 |0.123 |0.154 |0.142 |-0.11 |

|  |(1.292) |(1.369) |(1.349) |(1.369) |(1.373) |(1.491) |

|Share swap dummy |-1.325 |-2.185 |-1.227 |-1.785 |-1.753 |-2.183 |

| |(1.544) |(1.833) |(1.529) |(1.679) |(1.670) |(1.914) |

|Bankruptcy dummy |-0.731 |-1.188 |-0.327 |-0.775 |-0.784 |-0.589 |

|  |(1.572) |(1.756) |(1.706) |(1.784) |(1.785) |(1.792) |

|Reorganization dummy |-0.447 |-0.689 |0.054 |-0.282 |-0.307 |0.078 |

| |(1.655) |(1.773) |(1.744) |(1.773) |(1.782) |(1.889) |

|Asset stripping dummy |2.122 |2.300 |2.818* |2.712* |2.684* | |

|  |(1.471) |(1.456) |(1.506) |(1.435) |(1.435) | |

|Other for of dilution dummy |0.302 |0.095 |0.988 |0.731 |0.690 |0.152 |

| |(1.402) |(1.503) |(1.391) |(1.367) |(1.372) |(1.672) |

|Maximum loss due to |0.020* |0.023* |0.016 |0.019 |0.019 |0.024* |

|Dilution |(0.012) |(0.014) |(0.014) |(0.014) |(0.014) |(0.013) |

|Maximum loss due to |0.281 |0.431 |0.379 |0.493 |0.486 |0.349 |

|disenfranchisement |(0.547) |(0.609) |(0.596) |(0.612) |(0.612) |(0.606) |

|Observations |94 |94 |94 |94 |94 |93 |

Panel VB: Different types of coverage

| |I |II |

|Natural "newsworthiness" |0.452** |0.161 |

| |(0.219) |(0.270) |

|Log of (1+ # articles in Russian newspapers) |-0.017 | |

| |(0.359) | |

|Log of (1+ # articles in FT) | |0.653 |

| | |(0.739) |

|Log of (1+ # articles in WSJ) | |1.874** |

| | |(0.864) |

|Foreign ownership (%) |0.163 |-0.795 |

|  |(1.884) |(2.058) |

|EBRD as an investor dummy |1.131* |1.302* |

| |(0.616) |(0.704) |

|Log of assets |-0.02 |-0.172 |

|  |(0.160) |(0.162) |

|Dummy for oil industry |-0.419 |-0.773 |

| |(0.764) |(0.781) |

|Controls for nature of violation |

|Share issuance dummy |-0.125 |0.202 |

|  |(1.318) |(1.397) |

|Share swap dummy |-1.356 |-1.855 |

| |(1.585) |(1.678) |

|Bankruptcy dummy |-0.835 |-0.759 |

|  |(1.616) |(1.764) |

|Reorganization dummy |-0.656 |-0.514 |

| |(1.725) |(1.835) |

|Asset stripping dummy |2.022 |2.745* |

|  |(1.448) |(1.492) |

|Other form of dilution dummy |0.058 |0.502 |

| |(1.423) |(1.388) |

|Maximum loss due to dilution |0.021 |0.021 |

|  |(0.013) |(0.013) |

|Maximum loss due to disenfranchisement |0.291 |0.483 |

| |(0.607) |(0.625) |

|Observations |94 |94 |

Panel VC: Audience vs. Credibility

| |I |II |III |

|Natural "newsworthiness" |0.443* |0.416 |0.091 |

| |(0.227) |(0.264) |(0.255) |

|Log of (1+ # articles in Vedemosti) |-0.142 |-0.19 |-0.239 |

| |(0.510) |(0.540) |(0.522) |

|Log of (1+ # articles in other Russian newspapers) | |0.117 |-0.787 |

| | |(0.360) |(0.703) |

|Log of (1+ # articles in FT or WSJ) | | |2.318** |

| | | |(0.928) |

|Foreign ownership (%) |0.056 |-0.107 |-0.480 |

|  |(1.888) |(1.913) |(2.218) |

|EBRD as an investor dummy |1.134* |1.130* |1.173* |

| |(0.612) |(0.618) |(0.656) |

|Log of assets |-0.009 |-0.018 |-0.153 |

|  |(0.160) |(0.160) |(0.157) |

|Dummy for oil industry |-0.443 |-0.44 |-1.243 |

| |(0.769) |(0.766) |(0.924) |

|Controls for nature of violation |

|Share issuance dummy |-0.163 |-0.154 |0.167 |

|  |(1.303) |(1.301) |(1.326) |

|Share swap dummy |-1.361 |-1.383 |-1.536 |

| |(1.562) |(1.559) |(1.661) |

|Bankruptcy dummy |-0.876 |-0.892 |-0.674 |

|  |(1.609) |(1.615) |(1.782) |

|Reorganization dummy |-0.689 |-0.701 |-0.074 |

| |(1.734) |(1.745) |(1.781) |

|Asset stripping dummy |2.017 |2.090 |2.551* |

|  |(1.452) |(1.426) |(1.545) |

|Other form of dilution dummy |0.064 |0.120 |0.758 |

| |(1.452) |(1.446) |(1.446) |

|Maximum loss due to dilution |0.021* |0.021 |0.024 |

|  |(0.013) |(0.013) |(0.016) |

|Maximum loss due to disenfranchisement |0.296 |0.270 |0.849 |

| |(0.584) |(0.601) |(0.681) |

|Observations |94 |94 |94 |

Table VI: The Instrumental Variable Estimates

In this table we again look at outcome as the dependent variable. Outcome is defined to be equal to 0 if the potential governance violation was not redressed at all, 1 if partially redressed and 2 if substantially redressed. Column I is estimated by OLS. Column II is estimated by Instrumental Variables (IV), where the instrument for the log of the number of articles in the FT and WSJ is the percentage of Hermitage portfolio invested in a company at the end of 1998. Newsworthiness is measured as the log of one plus the number of references to a company in the WSJ and FT in the 6 month period from January to end of June 1998. We define control variables in Table 3. All the estimates are obtained by OLS. Huber/ White robust standard errors are reported in brackets, * means significant at 10%, ** 5% and *** 1%.

| |OLS |IV |

|Log of (1+ # articles in FT or WSJ) |0.438*** |0.855** |

| |(0.148) |(0.432) |

|Natural "newsworthiness" |0.051 |-0.053 |

| |(0.071) |(0.133) |

|Foreign ownership (%) |-0.111 |-0.342 |

|  |(0.567) |(0.609) |

|EBRD as an investor dummy |0.288 |0.275 |

| |(0.221) |(0.248) |

|Log of assets |-0.048 |-0.084 |

|  |(0.043) |(0.059) |

|Dummy for oil industry |-0.155 |-0.234 |

| |(0.187) |(0.190) |

|Controls for nature of violation |

|Share issuance dummy |0.014 |0.039 |

|  |(0.324) |(0.331) |

|Share swap dummy |-0.416 |-0.462 |

| |(0.336) |(0.416) |

|Bankruptcy dummy |-0.303 |-0.291 |

|  |(0.392) |(0.414) |

|Reorganization dummy |-0.192 |-0.121 |

| |(0.435) |(0.441) |

|Asset stripping dummy |0.760* |0.875** |

|  |(0.387) |(0.395) |

|Other form of dilution dummy |0.104 |0.239 |

| |(0.318) |(0.333) |

|Maximum loss due to dilution |0.005 |0.005 |

|  |(0.003) |(0.004) |

|Maximum loss due to disenfranchisement |0.097 |0.126 |

| |(0.136) |(0.135) |

|Observations |94 |94 |

|R-squared |0.22 |0.148 |

Table VII: Hazard Estimates of the Probability of Reversal

This tables uses a cox proportional hazard model to examine whether news coverage influences the timing of reversal of a proposed governance violation. Duration is the time between first public recognition of the proposed violation and resolution (partial or complete), if there was resolution within 12 months. Our explanatory variable is the cumulative number of news stories that mention the violation (not including possible stories about the resolution) from announcement until that month. Column 1 includes as resolutions those observations where the resolution occurred within 6 months from announcement. Column 2 includes in addition resolutions between 6 months and a year, with this time period collapsed as one additional observation. Robust standard errors are in brackets. *** indicates significant at 1% level.

.

| |Resolution |Resolution |

| |within the |within first 12|

| |first 6 months |months |

|Cumulative number of stories in Anglo-American Press |1.119*** |1.096*** |

| |(0.026) |(0.022) |

|Chi 2 |24.5 |20.4 |

|Prob > Chi 2 |0.00 |0.00 |

| | | |

|Number of observations |560 |560 |

|Number of subjects |98 |98 |

|Number of failures |21 |28 |

Table VIII: Who Acts to Redress Reported Corporate Governance Violations?

In this table we classify the successful outcomes according to the primary mechanism through which media pressure worked. Coding of the primary mechanism is based upon a reading of the Russian and English language press.

| |Intervention by Government Actors |Private Sector Actors |

| |Intervention of |Political intervention |Company relents when |Company relents |

| |regulators/courts | |faced with significant | |

| | | |opposition | |

| |AvtoVAZ |Gazprom |KAMAZ |Chernogorneft |

| |Irkutskelectrosvyaz |RAO UES |KrAZ |GAZ |

| |Kuznetsk Ferrous Alloys|Sberbank |Ust-Ilimsk Timber |Gazprom |

| |Tomsk Refinery |Novoship |Viksunsk Pipe |Transneft |

| |Tomskneft | |Bratsk Pulp |Rosshelf |

| |Yuganskneftegaz | |KomiTEK |Bashkirenergo |

| |Sidanco | |Volgotanker | |

| |Lomonosov Porcelain | |VSMPO | |

| |Elektrosila | | | |

| |AVISMA | | | |

|Number of cases |10 |4 |8 |6 |

|Percentage of |36% |14% |29% |21% |

|cases with | | | | |

|positive outcome | | | | |

-----------------------

[1] Alan Murray “Hedge Funds Are New Sheriffs of Boardroom,” Wall Street Journal, 14 December 2005, pg A2.

[2] When discussing governance violations we focus on the distributional impact. It is harder to make any overall welfare assessment. Even actions that have an extremely negative distributional impact (such as pure theft) can have a positive efficiency effect, because the consolidation of cash flow rights in one hand can have positive incentive effects as argued in the Russian case by Shleifer (2005), and Guriev and Rachinsky (2005).

[3] These data are based on the calculation .05*[.57*book or market value of equity in dollars in January 2002]. We compute book value of equity based on the 80 companies that remain alive this date, and market values for 26 companies that remain alive in 2002 where there is sufficient liquidity in traded stock to compute a meaningful market value.

[4]

[5] Louis D. Brandeis, 1933, Other People’s Money, National Home Library Foundation: 62.

[6] The large literature on law and finance has emphasized the importance of legal enforcement as different from the law on the books (La Porta et al. , 1998; Bhattacharya and Daouk, 2002), but has not explored what drives enforcement. As the discussion below suggests, media pressure can be an important determinant of legal enforcement.

[7] This is not strictly true with a regulatory agency such as the SEC where those in charge have no voters to be accountable to. But it is a reasonable approximation, for the SEC relies for its budget and authority on Congress, and these political overseers care about political concerns about inactivity.

[8] “Oily Charm”, The Economist, 5 December 1998.

[9] Yukos Press release as reported by PR Newswire, 30 August 1999, 03:37 PM.

[10] Robert Cottrell and Arkady Ostrovsky: “After the oligarchs,” Financial Times, 16 April 2001.

[11] One way to reconcile this effect with the traditional reputation effect is to posit that every manager has the option to start a political career and, thus, she cares about her general reputation. But since a political career is not a source of large monetary gains, the existence of such an interest can be justified only with an extra term in the utility function. It is simpler, then, to posit from the beginning that managers care about their reputation in this broader sense.

[12] The Data Appendix on the Journal of Finance website provides further information on the various lawsuits brought by Hermitage and the companies involved

[13] The appendix on the Journal website identifies teh type of alleged violation for each event in our sample.

[14] As Mikhail Lesin, formerly President Putin’s Media Minister, described it, Russian language media are characterized by ““information wars,” which saw oligarchic groups trying to destroy each other through the media, ultimately causing the media’s authority to dwindle and undermining trust in the written word.” February 24, 2005, “Protection and the Media”, .

[15] In an unreported table we also look at the cross correlation between these variables. The highest level of correlation is -0.65 (between the two measures of dilution). The only other one above 0.5 is between log assets and log news (0.52).

[16] We also explored an earlier period, July 1, 1997 to December 31, 1997 and found a very high correlation (.966) between the two measures and that our results are strengthened with this alternative measure.

[17] This is the earliest financial statement we have available. For the 5 companies with an event before December 1998 we use the December 1998 holdings unless we know from the financial statement when the stake was acquired or disposed. We checked that our results are robust to dropping these 5 observations.

[18] If we drop these dummies the results are substantially unchanged.

[19] For four observations the violation involves assets where we were unable to assess their market value so were unable to calculate the extent of the dilution.

[20] We use this longer window to capture the possibility that there might be noise in our identification of the announcement date (e.g. the bulletin on corporate governance actions is issued only weekly). All results are robust to focusing on the narrower window of t to 2 months after the announcement of the proposed infraction.

[21] For simplicity we have presented results using one set of controls. The results are robust to alternative specifications including omitting the severity of infraction variable (which allows us to use all 98 observations), excluding the category of violation, using a more comprehensive set of industry controls rather than a simple oil dummy, using another measure of foreign interest in a stock (Troika Dialog provides company analysis including governance scores) to address possible concerns about mismeasurement of foreign ownership..

[22] We also checked this with an alternative period, July 1, 1997 to December 31, 1997 (correlation =.966 with later period) and found our results to be robust.

[23] We thank the referee for this comment.

[24] We thank the referee for this suggestion.

[25] Since we do not have the detailed timing of the article publication after 6 months, we collapse the last 6 months into one period.

[26] Vasiliev resigned from FSCR in October 1999 complaining about a lack of support to address governance violations.

[27] Take for example, Wal Mart. Until recently it paid little attention to PR and political lobbying. . After some major political setbacks, mainly due to Union activism, Wal Mart reversed its policy.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download